AI crap - Why ML will make the world worse, not better
drewdevault.com
There is a machine learning bubble, but the technology is here to stay. Once the bubble pops, the world will be changed by machine learning. But it will probably be crappier, not better.
What will happen to AI is boring old capitalism. Its staying power will come in the form of replacing competent, expensive humans with crappy, cheap robots.
AI is defined by aggressive capitalism. The hype bubble has been engineered by investors and capitalists dumping money into it, and the returns they expect on that investment are going to come out of your pocket. The singularity is not coming, but the most realistic promises of AI are going to make the world worse. The AI revolution is here, and I don’t really like it.
You could have said the same for factories in the 18th century. But instead of the reactionary sentiment to just reject the new, we should be pushing for ways to have it work for everyone.
I don't see how rejecting 18th century-style factories or exploitative neural networks is a bad thing. We should have the option of saying "no" to the ideas of capitalists looking for a quick buck. There was an insightful blog post that I can't find right now...
Lets not forget all the exploitation that happened in that period also. People, even children, working for endless hours for nearly no pay, losing limbs to machinery and simply getting discarded for it. Just as there is a history of technology, there is a history of it being used inequitably and even sociopathically, through greed that has no consideration for human well-being. It took a lot of fighting, often literally, to get to the point we have some dignity, and even that is being eroded.
I get your point, it's not the tech, it's the system, and while I lost all excitement for AI I don't think that genie can't be put back in the bottle. But if the whole system isn't changing, we should at least regulate the tech.
But AI will eliminate so many jobs that it will affect a lot of people, and strain the whole system even more. There isn't a "just become a programmer" solution to AI, because even intellectually-oriented jobs are now on the line for elimination. This won't create more jobs than it takes away.
Which shows why people are so fearful of this tech. Freeing people from manual labor to go to intellectual work was overall good, though in retrospect even then it came at a cost of passionate artisans. But now people might be "freed" from being artists to having to become sweatshop workers, who can't outperform machines so their only option is to undercut them. Who is being helped by this?
Yes, I know about the exploitation that happened during early industrialization, and it was horrible. But if people had just rejected and banned factories back then, we'd still be living in feudalism.
I know that I don't want to work a job that can be easily automated, but intentionally isn't just so I can "have a purpose".
What will happen if AI were to automate all jobs? In the most extreme case, where literally everyone lost their job, then nobody would be able to buy stuff, but also, no company would be able to sell products and make profit. Then, either capitalism would collapse - or more likely, it will adapt by implementing some mechanism such as UBI. Of course, the real effect of AI will not be quite that extreme, but it may well destabilize things.
That said, if you want to change the system, it's exactly in periods of instability that can be done. So I'm not going to try to stop progress and cling to the status quo out of fear what those changes might be - and instead join a movement that tries to shape them.
Maybe. But generally on Lemmy I see sooo many articles about "Oh, no, AI bad". But no good suggestions on what exactly regulations should we want.
Movements that shape changes can also happen by resisting or by popular pressure. There is no lack of well-reasoned articles about the issues with AI and how they should be addressed, or even how they should have been addressed before AI engineers charged ahead not even asking for forgiveness after also not asking for permission. The thing is that AI proponents and the companies embracing them don't care to listen, and governments are infamously slow to act.
For all that is said of "progress", a word with a misleading connotation, once again this technology puts wealthy people, who can build data centers for it, at an advantage compared to regular people who at best can only use lesser versions of it, if even that, they might instead just receive the end result of whatever the technology owners want to offer. Like the article itself mentions, it has immense potential for advertising, scams and political propaganda. I haven't seen AI proponents offering meaningful rebuttals to that.
At this point I'm bracing for the dystopian horrors that will come before it all comes to a head, and who knows how it might turn out this time around.
You won't get a direct rebuttal because, obviously, an AI can be used to write ads, scams and political propaganda.
But every day millions of people are cut by knives. It hurts. A lot. Sometimes the injuries are fatal. Does that mean knives are evil and ruining the world? I'd argue not. I love my kitchen knives and couldn't imagine doing without them.
I'd also argue LLMs can be used fact check and uncover scams/political propaganda/etc and can lower the cost of content production to the point where you don't need awful advertisements to cover the production costs.
This knife argument is overused as an excuse to take no precautions about anything whatsoever. The tech industry could stand to be more responsible about what it makes rather than shrugging it off until aging politicians realize this needs to be adressed.
Using LLMs to fact check is a flawed proposition, because ultimately what it provides are language patterns, not verified information. Nevermind its many examples of mistakes, it's very easy for them to provide incorrect answers that are widely repeated misconceptions. You may not blame the LLM for that, you can scratch that to generalized ignorance, but it still ends up falling short for this use case.
But as much as I dislike ads, that last one is part of the problem. Humans losing their livelihood. So, going back to a previous point, how does the lowered ad budget help anyone but executives and investors? The former ad workers get freed to do what? Because the ones focused on art or writing would only have a harder time making a career out of that now.
Accelerationists?
Everyone who died as a result of their introduction probably would say the same, yes. If corpses could speak, anyway.
Well if you can find anyone who's died because an AI wrote an article then I'll concede you have a point.
Did you read the whole article including the "flame bait"? The author gives an example there of someone committing suicide because an AI encouraged them...
Is that the AI's fault, or the depressed and suicidal human's fault?
Do you not think that the person would have committed suicide whether they asked the AI or not? The AI might have sped up the decision, but it is the human who made it.
It is not like the AI is out there trying to convince non-depressed humans to become depressed in order they go kill themselves ...
Well, in the linked article the wife of this person said that they wouldn't have committed suicide without the AI facilitating it. So yes, I would say it is at least in part the AI's fault. And no, I didn't say it was the intention of the AI to do so. But that doesn't mean it won't do it at all.
You seem to really wanna push AI and lose your empathy over this...
If the technology actually existed to replace human workers, the human workers could chip in and buy the means of production and replace the company owners as well.
If the humans are replaced, how will they afford to buy what the company owners aren't selling?
Top quality luddite opinions right here. Plenty of fear and oprobium being directed against the technology, while taking the kleprocratic capitalism and kakistocracy as a given that can't be challenged.
That seems to be the theme of the era.
Yes, it is incompatible with the status quo. That's a good thing. The status quo is unsustainable. The status quo is on course to kill us all.
The only real danger AI brings is it will let our current corrupt leaders and corrupt institutions be more efficient in their corruption. The problem there is not the AI; it's the corruption.
Improving human efficiency is essentially the purpose of technology after all. Any new invention will generally have this effect.
But there's also real danger here today.
For example, ‘Life or Death:’ AI-Generated Mushroom Foraging Books Are All Over Amazon
These are easily avoidable problems. There are always reputable authors on topics and why would a self published foraging book by some random person be better than an AI one? You buy books written by experts, especially when it’s about life or death.
"Easily avoidable" if you know to look for them or if they're labelled appropriately. This was just an example of a danger that autocomplete AI is creating today. Unscrupulous people will continue to shit out AI generated nonsense to try to sell when the seller does zero vetting of the products in their store (one of the many reasons I no longer shop at Amazon).
Many people, especially beginners, are not going to take the time to fully investigate their sources of knowledge, and to be honest they probably shouldn't have to. If you get a book about mushrooms from the library, you can probably assume it's giving valid information as the library has people to vet books. People will see Amazon as being responsible for keeping them safe, for better or worse.
I agree that generally there is a bunch of nonsense about ChatGPT and LLM AIs that isn't really valid, and we're seeing some amount of AI bubble happening where it's a self feeding thing. In the end it will shake out, but before that all happens you have some outright dangerous and harmful things occurring today.
I mean, people should at least check if the publisher is reliable for any information source.
I think the idea is that someone buying a basic book on foraging mushrooms isn't going to know who the experts are.
They're going to google it, and they're going to find AI-generated reviews (with affiliate links!) of AI-generated foraging books.
Now, if said AI is generating foraging books more accurate than humans, that's fine by me. Until that's the case, we should be marking AI-generated books in some clear way.
The problem is, the LLM AIs we have today literally cannot do this because they are not thinking machines. These AIs are beefed-up autocompletes without any actual knowledge of the underlying information being conveyed. The sentences are grammatically correct and read (mostly) like we would expect human written words to read, however the actual factual content is non-existent. The appearance of correctness just comes from the fact that the model was trained on information that was (probably mostly) correct in the first place.
I mean, we should still be calling these things algorithms and not "AI" as "AI" carries a lot of subtext in people's minds. Most people understand "algorithms" to mean math, and that dehumanizes it. If you call something AI, all of a sudden people have sci-fi ideas of truly independent thinking machines. ChatGPT is not that, at all.
I agree. And ML may never be able to cross that line.
That said, we've been calling it AI for decades now. It was weird enough to me when people started using ML more. I remember the AI classes I took in college, and the AI experts I met in my jobs. Then one day it was "just ML". In most situations, it's the same darn thing.
It's literally baked into the models themselves. AI will reinforce kleptocratic capitalism and kakistocracy as you so aptly put it because the very data it's trained on is a slice of the society it resembles. People on the internet share bad, racist opinions and the bots trained on this data do the same. When AI models are put in charge of systems because it's cheaper than putting humans in place, the systems themselves become entrenched in status-quo. The problem isn't so much the technology itself, but how the technology is being rolled out, driven by capitalistic incentives, and the consequences that brings.
snicker drewdevault is an avid critic of capitalism. thats entirely the point of this post actually.
Then it is horrifically badly written. Maybe get an AI to give it a once over?
I'm getting so so tired of these "AI/ML bad, world is doom" articles being posted multiple times a day. who is funding these narratives??
you have no idea who drew devault is clearly and you entirely missed the point of what he posted.
I also have no idea who he is and I also missed the point. It's just another "AI bad" article, even if the message this time is "AI bad, but not as bad as you think."
hes pointing out how the technology is going to be used realistically by corporations. hes not saying AI is bad inherently. hes saying the outcome will be bad for society. which is very true.
And its a hot topic so there are a lot of articles about it since it generates traffic. And the bigger and doomier the article, the more hype.
But lots of people are unhappy about AI, its not a narrative being funded by a secret cabal or something.
Sure, it's all about capitalism. Nothing good like this could ever come from advances in technology:
https://www.ucsf.edu/news/2023/08/425986/how-artificial-intelligence-gave-paralyzed-woman-her-voice-back
ML is a tool and like most tools it has broad use cases. Some of them are very, very, good.
There is a name for this debating technique where you go "sure, there was nothing good about Hitler - except he cared about dogs!". Can't remember. Is it strawman?
I think we all understand that capitalism is mostly bad for humans, and really good for corporations and their owners. AI and robots will be exploited to replace people since they are massively more powerful and much cheaper.
A few things will be better I guess, but most will be worse. People already are not actually needed to work this much anymore, and as soon as they can be replaced with something cheaper and more efficient they will. That is capitalism.
A strawman argument is where you ignore what was said by the other person and instead respond with something distorted. That's not what I did - the core premise of Drew's argument is that AI will not "make the world better" and I provided a crystal clear example of how it makes the world better.
It was just one example, and obviously not the complete picture, but what choice do I have? It's such a broad topic I couldn't possibly list everything AI will impact without writing an entire book.
No I disagree. Corporations exist exclusively to benefit their human owners them. Which means anything that's "good for corporations" is good for a select small number of humans.
Don't blame "capitalism" for wealth inequality. Blame the actual humans (e.g. Donald Trump, Elon Musk) who have made it their life's work to drive the global economy even harder into a world that benefits the fiew and ignores the struggles of the many.
Also - not all corporations are bad. Some of them do great work that truly benefits the world and I would personally put OpenAI in that category. Their mandate is not to make a profit - and in fact the amount of profit they can legally make has been limited. Their mission is literally "to ensure that artificial general intelligence benefits all of humanity". I hope they succeed, and I think they will. Drew is wrong.
Who buys the stuff if everyone is replaced?
Eventually nobody.
Capitalism isn't about sustainability, it's about making the most amount of profit in the shortest amount of time.
Eventuall you bleed everyone dry and nobody has a job. But for a short amount of time the shareholders will have had a huge number of 0's and 1's in a database somewhere equating to their "worth"
Humans are consumers, they will buy stuff. But most won't work for corporations anymore since robots and AI are far more effective at most jobs.
Humans will still buy the stuff robots produce. Maybe the money will come from governments as some kind of citizen coins, distributed differently based on some criteria. Not sure.
That would be UBI and that’s seen as an improvement by a lot of people so why stand in the way of that? Robots do the work, people get a budget they can spend on that work while they don’t have to do it.
Because none of those people have policy changing influence. Nobody here is standing in the way of it, most people here are advocates for it. But we don't write the laws, and we only get to vote on a very small percentage of them.
I disagree. If we replace this writer with ChatGPT4, it would generate a more balanced article.
More balanced articles are not necessarily better though. I'd dather read two conflicting opinions that are well thought out than a mild compromise with unknown bias.
That's where it all falls down of course. Because these opinions are anything but well thought out.
Why would ChatGPT be more “balanced,” what does “balanced” mean, and why is it better?
More balanced than what?
ChatGPT ingest lots of articles from the web and newspapers, identify patterns in the text, and generate relevant reply based on what it ingested.
I expect ChatGPT to perpetuate biases found in its training data, and don't see how it'd improve balance.
What I mean is that the article was full of negative bias.
ChatGPT 4, when used with care, can take into account different opinions, both positive and negative.
Unlikely to replace the "most" competent humans, but probably the lower 80% (Pareto principle), where "crappy" is "good enough".
What's really troubling, is that it will happen all across the board; I'm yet to find a single field where most tasks couldn't be replaced by an AI. Used to think 3D design would take the longest, but no, there are already 3D design AIs.
Critical-application development. For example, developing a program that drives a rocket or an airplane.
You can have an AI write some code. But good luck proving that the code meets all the safety criteria.
You just said the same thing the comment responding to did, though. He pointed out that AI can replace the lower 80%, and you said the AI can write some code but that it might have trouble doing the expert work of proving the code meets the safety criteria. That's where the 20% comes in.
Also, it becomes easier to recognize the possibility for AI contribution when you widen your view to consider all the work required for critical application development beyond just the particular task of writing code. The company surrounding that task has a lot of non-coding work that gets done that is also amenable to AI replacement.
That split won't work cause the top 20% would not like to do their day job clean up AI codes. It's much better time investment wise for them to write their own template generation tool so the 80% can write the key part of their task, than taking AI templates that may or may not be wrong and then hunting all over the place to remove bugs.
Use the AI to fix the bugs.
A couple months ago, I tried it on ChatGPT: I had never ever written or seen a single line in COBOL... so I asked ChatGPT to write me a program to print the first 10 elements of the Fibonacci series. I copy+pasted it into a COBOL web emulator... and it failed, with some errors. Copy+pasted the errors back to ChatGPT, asked it to fix them, and at the second or third iteration, the program was working as intended.
If an AI were to run with enough context to keep all the requirements for a module, then iterate with input from a test suite, all one would need to write would be the requirements. Use the AI to also write the tests for each requirement, maybe make a library of them, and the core development loop could be reduced to ticking boxes for the requirements you wanted for each module... but maybe an AI could do that too?
Weird times are coming. 😐
I'm a professional programmer and this is how I use ChatGPT. Instead of asking it "give me a script to do big complicated task" and then laughing at it when it fails, I tell it "give me a script to do ." Then when I confirm that works, I say "okay, now add a function that takes the output of the first function and does " Repeat until done, correcting it when it makes mistakes. You still need to know how to spot problems but it's way faster than writing it myself, even if I don't have to go rummaging through API documentation and whatnot.
I mean that is exactly what programming is except you type to an AI and have it type the script. What is that good for?
Could have just typed the script in the first place.
It ChatGPT can use the API it can't be too complex otherwise you are in for a surprise once you find out what ChatGPT didn't care about (caching, usage limits, pricing, usage contracts)
Sure - but ChatGPT can type faster than me. And for simple tasks, CoPilot is even faster.
Also - it doesn't just speed up typing, it also speeds up basics like "what did bob name that function?"
And stuff like "I know there's a library out there that does the thing I'm trying to do, what's it named and how do I call it?"
I haven't been using ChatGPT for the "meat" of my programming, but there are so many things that little one-off scrappy Python scripts make so much easier in my line of work.
I already explained.
I could write the scripts myself, sure. But can I write the scripts in a matter of minutes? Even with a bit of debugging time thrown in, and the time it takes to describe the problem to ChatGPT, it's not even close. And those descriptions of the problem make for good documentation to boot.
Fashion designers are being replaced by AI.
Investment capitalists are starting to argue that C-Suite company officers are costing companies too much money.
Our Ouroboros economy hungers.
C-Suites can get replaced by AIs... controlled by a crypto DAO replacing the board. And now that we're at it, replace all workers by AIs, and investors by AI trading bots.
Why have any humans, when you can put in some initial capital, and have the bot invert in a DAO that controls a full-AI company. Bonus points if all the clients are also AIs.
The future is going to be weird AF. 😆😰🙈
If the AI is doing a better job at each of those things, why not let it?
That's where we need to ask how we define "better". Is better "when the number goes bigger" or is better "when more people benefit"? If an AI can better optimize to better extract the maximum value from people's work and discard them, then optimize how many ways they can monetize their product to maximize the profit they get from each customer, the result is a horrible company and a horrible society.
In theory yes... but what do we call "doing a better job"? Is it just blindly extracting money? Or is it something more, and do we all agree on what it is? I think there could be a compounded problem of oversight.
Like, right now an employee pays/invests some money into a retirement fund, whose managers invest into several mutual funds, whose managers invest into several companies, whose owners ask for some performance from their C-suite, who through a chain of command tell the same employee what to do. Even though it's part of the employee's capital that's controlling that company, if it takes an action negative for the employee like fracking under their home, or firing them, they're powerless to do anything about it with their investment.
With AI replacing all those steps, it would all happen much quicker, and —since AIs are still basically a black box— with even less transparency than having corruptible humans on the same steps (at least we kind of know what tends to corrupt humans). Adding strict "code as contract" rules to try to keep them in check, would on a first sight look like an improvement, but in practice any unpredicted behavior could spread blindingly fast over the whole ecosystem, with nobody having a "stop" button anymore. That's even before considering coding errors and malicious actors.
I guess a possible solution would be requiring every AI to have an external stop trigger, that a judicial system could launch to... possibly paralyze the whole economy. But that would require new legislation to be passed (with AI lawyers), and it would likely get late, and not be fully implemented by those trying to outsmart the system. Replace the judges by AIs too, politicians with AIs, talking heads on TV with AIs... and it becomes an AI world where humans have little to nothing to say. Are humans even of any use, in such a world?
None of those AIs need to be an AGI, so we could run ourselves into a corner with nobody and nothing having a global plan or oversight. Kind of like right now, but worse for the people.
Alternatively, all those AIs could be eco-friendly humans-first compassionate black boxes... but I kind of doubt those are the kind of AIs that current businesses are trying to build.
Thing is nobody will do that because once AI finds a way to spazz out that is totally unpredictable (black box) everything might just be gone.
It's a totally unrealistic scenario.
They are working on mitigating the unpredictable "black box".
Like making the AI explain their working method step by step. Not only does it make the AI more transparent, it also increases the correctness of whatever it types.
AI is still in development. It is good to list the problems you have, but don't think those problems won't be solved in the future.
People are already doing it, piece by piece, in all areas. As more AIs get input from other AIs, the chance of a cascading failure increases... but it will seem to be working "good enough" up until then, so more people will keep jumping on the bandwagon.
The question is: can we prepare for the eventual cascading spazz out, or have we no option other than letting it catch us by surprise?
Unfortunately everything AI does is kind of shitty. Sure you might have a query for which the chosen AI works well but you might as well not.
It you accept that it sometimes just doesn't work at all sure AI is your revolution. Unfortunately there are not too many use cases where this is helpful.
I posit that in 80% of the cases, an AI working well even less than 50% of the times, is still "good enough" to achieve the shittier 80% of goals.
"I'll have a burger with extra ketchup"... you get extra mayo instead... for half the price; "good enough".
This is my worry as well